Abstract The experience of drug craving is known to increase drug use1-6 and relapse after treatment7-16. Specifically, drug-related stimuli are known to elicit cue-induced craving, as well as physiological and neural responses, called cue reactivity, which have all been directly linked to drug use and relapse across addictions17-35, including gambling36-39. To date, hundreds of studies investigated neural cue-reactivity; however, there is little consensus in findings, due to significant variation in methods, drug types, cue types, and analysis procedures. Further, prior meta-analyses used suboptimal methods and inclusion criteria, and were underpowered40-44. In Aim 1, we propose to use state-of-the-art meta-analytic methods45-47 to summarize findings from ~120 published imaging studies, representing >3000 participants during neural cue reactivity/cue- induced craving (and 3x larger than any prior meta-analysis). Following our prior meta-analytic work in other domains48-61, we will use optimized inclusion criteria and correction methods, and account for multiple methodological differences in studies across substances (e.g., cigarettes, alcohol, cocaine, cannabis) and gambling, a behavioral addiction. Further, we will test for differences across drug types (e.g., cigarettes vs. alcohol) and different cue-types (e.g., pictures vs. video) to resolve several open questions in the field (e.g., role of insula in cigarette craving vs. drug craving in general62, 63; pre-scan abstinence64, 65). In Aim 2, we will go beyond meta-analysis to address an urgent need in addiction research66: development of predictive biomarkers67-69, which are stable indicators of biological processes. Biomarkers have been developed in other areas of medicine70, and we have recently done this in pain71, negative emotion72-74, and empathy74. To do this, we will combine the meta-analytic results obtained in Aim 1 with person-level fMRI data from our lab (of cue-induced craving75-79) using machine learning models, to establish a multivariate pattern of neural activity that can predict craving self-report from neural activity ? a predictive neuromarker for craving. Given the associations between neural cue reactivity, craving, and drug use outcome, a multivariate neuromarker could provide a powerful neural predictor of outcomes66, 67. Thus, in Aim 3, we will validate this neuromarker on data from two ongoing clinical trials of N=128 cigarette smokers and N=150 cocaine users collected at Yale's Psychotherapy Development Center (P50 DA09241; Center PI: Carroll; Project PI: Kober). Here, the neuromarker will be used on pre-treatment fMRI cue-reactivity data to predict both self-reported craving and long-term clinical outcomes (e.g., cocaine abstinence). The validated neuromarker will provide well-defined brain targets that can then be used to test the efficacy of interventions to reduce craving and drug use, such as pharmacotherapies, brain stimulation, or cognitive training, and to monitor patient progress in both research and clinical settings. !